Lody is open sourcing its CLI and local desktop client under an Apache 2.0 license. It's not a new coding agent; it's a shared workspace layer that wraps the agents you already have. Claude Code, Codex, Kimi, OpenCode, and more can plug straight in. Teams can share sessions, watch running status and code changes, and dispatch tasks from a phone or web browser to a desktop or server.
When multiple agents are working at once, different sessions can be placed in separate Git worktrees to keep agents from colliding on code changes. Agents can also create, read, and resume other agents' sessions. A lead agent can split debugging, implementation, and testing across child sessions running in parallel.
That doesn't mean Lody's entire stack is now open source. The public repo explicitly leaves out the hosted backend, deployment configs, billing system, and the web and mobile app source code. Cross-device collaboration still depends on Lody's hosted sync service, and end-to-end encryption isn't supported yet. The team's next move is to lean further into local-first, letting sessions, docs, tasks, and historical decisions become project context that teams own themselves.
https://twitter.com/lody_ai/status/2092605698773516328